MVS TCP/IP 3.3

At 04:17 PM 05/12/97 -0400, Jerry Lawson wrote:

>Date:     May 12, 1997             Time: 1:55 PM
>From:     Jerry Lawson
>          The Hartford Insurance Group
>          (860) 547-2960           jlawson AT thehartford DOT com
>
>For the last couple of weeks there has been discussion about MVS TCP/IP 3.2,
>and the problems it has been causing.  I was talking with the Network folks
>here, and the lady who supports TCP/IP related other problems that they had
>been having with 3.2 - apparently it's not a pretty sight.
>
>Then she asked me to post a question to the list - is anyone out there using
>MVS TCP/IP 3.3 yet?  If so, is it any better?
>
>                                         Jerry
>

----
Hi Jerry, all.
Hi Jerry, all.

My understanding is that IBM TCP/IP 3.3.0 is essentially an
OpenEdition only release. The 3.3.0 stack runs as an OpenEdition
application and the socket interface supplied with 3.3.0 can
only be invoked from OpenEdition (read: UNIX) applications.
ADSM and other "native" MVS applications will not be able
to use the TCP/IP 3.3.0 facilities (unless they explicitly become
OpenEdition enabled...).

If you read between the lines in the IBM announcement letters
for TCP/IP you can deduce that, at least for the interim,
you must run TCP/IP 3.2.0 and TCP/IP 3.3.0 concurrently,
using 3.2.0 for native MVS applications and 3.3.0 for
OpenEdition (UNIX) Applications. It's a bit more of a
stretch, but in my opinion you can also deduce that TCP/IP 3.4.0
(or whatever...) will be strictly an OpenEdition based TCP/IP stack,
but will also provide an upward compatible sockets interface
for native MVS applications (allowing native MVS applications
to directly access the 3.4.0 OpenEdition based stack).

All this might seem a bit confusing, but the bottom line here
is that native MVS applications will use TCP/IP 3.2.0 until
TCP/IP 3.4.0 is released, whereas OpenEdition applications can
use either TCP/IP 3.3.0 or TCP/IP 3.4.0.

One reason that I feel TCP/IP 3.3.0 was needed is that OpenEdition
(i.e. UNIX) applications often rely heavily on intra-host TCP/IP
communication, and having to use the "native" TCP/IP 3.2.0 stack
could not provide the performance needed by the OpenEdition platform.
So, IBM needed to deliver (and rather quickly) a TCP/IP that could
adequately service the growing needs of OpenEdition applications.
Hence the release of TCP/IP 3.3.0. I would also suspect that having a
OpenEdition based TCP/IP stack makes it much easier for IBM to keep up
with RFC's and the like. It's always been easier to port UNIX
applications, including TCP/IP, to other UNIX platforms (OpenEdition),
rather than to native MVS. Beginning with TCP/IP 3.4.0, IBM should be
able to support a single UNIX like implementation of TCP/IP, for all
System/390 based applications.

In case you haven't already figured it out, TCP/IP 3.4.0 will
*REQUIRE* that OpenEdition be installed, configured, and available
on your MVS, excuse me, OS/390 platform. TCP/IP 3.4.0 is not a
native MVS application like TCP/IP 3.2.0, it is a very very UNIX
implementation of TCP/IP (can you say HFS, VI, etc.?).
OpenEdition is coming whether you're ready or not!

One final note, our native MVS applications have been running
well with TCP/IP 3.2.0, both with the 3.1.0 sockets interface
and with the 3.2.0 High Performance Native Sockets (HPNS) interface.
Migrating to TCP/IP 3.2.0 was a lot *LESS* painful than we thought it
was going to be.

Regards,
Doug

P.S. The above is based on my personal interpretation of IBM
     information as an experienced TCP/IP application developer.
     If anyone else has a different point of view, I'm listening...

----
Doug Babcock <babcock AT tmd DOT com>
Doug Babcock <babcock AT tmd DOT com>
TMD Consulting Inc., a tiny little division of StorageTek
Standard disclaimers apply (free copy on request)
---
To see what I/we do, visit the CAM/EBF home page at
To see what I/we do, visit the CAM/EBF home page at
http://www.stortek.com/StorageTek/software/cam/camhome.html